Eligibility

Why Your Pathology Lab's AI Software Is SR&ED-Eligible (And Your Lab Manager Doesn't Know It)

Canadian digital pathology companies like Pathcore and Ibex Medical Analytics build AI that reads tissue slides. The R&D behind that software qualifies for SR&ED — but most medical research teams never file.

Dr. Priya Sundaram · SR&ED Research Desk 2026-05-26 6 min read

The $14 billion slide-reading problem

In 2024, the Canadian digital pathology market was valued at approximately $1.2 billion, growing at 14% annually. Companies like Pathcore (Toronto), Ibex Medical Analytics (working with Canadian health networks), and various hospital-affiliated research labs are building AI systems that detect cancer markers, quantify immune response, and flag abnormal tissue patterns — all from digitized microscope slides.

The twist: most of these labs think of themselves as medical operations, not technology companies. Their engineering teams build software that reads slides, quantifies staining intensity, and compares patterns across thousands of cases. That software development is R&D. And most of it qualifies for SR&ED.

The gap: medical research teams have some of the most systematically documented R&D in Canada — peer review, controlled experiments, statistical validation — but they rarely connect that documentation to Canada's R&D tax credit program.

What qualifies: the technical uncertainty in medical AI

A digital pathology team develops a convolutional neural network to detect Gleason pattern variations in prostate biopsies. Published models exist — but they were trained on datasets from European and American labs with different staining protocols, scanner resolutions, and tissue preparation standards.

The Canadian lab's slides look different. The staining is more variable. The scanner has a different colour profile. The tissue handling creates artefacts that confuse the published model. The team systematically tests preprocessing approaches, augmentation strategies, and model architectures to achieve accuracy comparable to the published benchmarks on their specific data.

Qualifying R&D scenario

A research lab at a Canadian teaching hospital trains a multi-class classifier to identify five types of renal cell carcinoma from whole-slide images. Published approaches (ResNet-based classifiers) achieve 89% accuracy on the benchmark dataset but drop to 71% on the lab's internal validation set. The team experiments with four architecture modifications, documents why each one improves or fails on specific artefact types (folding, air bubbles, staining variation), and ultimately develops a two-stage approach that achieves 87% accuracy. The systematic investigation — including the failed attempts and the artefact characterization — is documented in the lab's existing research protocols and qualifies for SR&ED.

What doesn't qualify: the routine work that looks like R&D

Not everything a pathology lab does with software is R&D. The distinction matters for claim integrity:

Not eligible: Deploying an existing FDA-approved or Health Canada-licensed diagnostic algorithm to the lab's scanner. The implementation work is routine integration.
Eligible: Modifying that algorithm because the lab's specific scanner model and staining protocol produce image characteristics the approved model wasn't validated on, and systematically documenting the performance delta.
Not eligible: Standard quality control workflows, slide digitization at standard resolutions, and routine LIS (Laboratory Information System) integrations.
Eligible: Developing a novel quality metric or slide-scoring approach where no existing validation framework exists for the lab's specific patient population and preparation methods.

The medical trap: assuming that because the work serves a clinical purpose, it's automatically R&D. CRA evaluates the technical investigation, not the medical outcome. A model that saves lives but was built using documented, established methods is not R&D. A model that modestly improves workflow but required genuine experimentation to solve a technical obstacle is R&D.

How doctors can document R&D without becoming engineers

The good news: medical research teams already document most of what CRA needs. They just document it for publication, not for tax credits. The same protocols, validation studies, and statistical analyses that support a journal submission also support an SR&ED claim.

Research protocols that describe the technical objective, the hypothesis, the methodology, and the expected outcomes
Validation studies comparing new approaches against established baselines on specific datasets
Statistical analyses showing performance differences and confidence intervals
IRB (Research Ethics Board) applications that describe the technical rationale for the study
Lab notebooks or digital logs recording experimental iterations, parameter changes, and observed outcomes

The missing piece: connecting the research documentation to the SR&ED framework. A journal article says 'we achieved 87% accuracy.' An SR&ED narrative says 'we encountered a specific technical obstacle — published approaches failed on our staining protocol — and systematically investigated four modifications, documenting why each one addressed specific artefact types.' Same work, different framing.

This guide is for general information only. Medical research institutions should consult with qualified Canadian CPAs who specialize in SR&ED for health technology and clinical research claims. Institutional policies may impose additional compliance requirements. Learn more at sredy.io.

digital pathology medical AI clinical research Pathcore Ibex health tech

Turn this guide into a claim package

SREDY.IO walks you through eligibility, project narratives, supporting costs, and evidence so your CPA has a cleaner file to review.

Check eligibility → View sample package

More guides

Eligibility

Check eligibility — free →